JP6016552B2 - Antireflection system - Google Patents

Description

   The present invention relates generally to image processing, and in particular to removing unwanted features from an image. More specifically, the present invention relates to a method and apparatus for removing unwanted bright spots in an image.

   Depending on the situation, undesired features may appear in the image of the scene generated using the sensor system. For example, light reflections from surfaces in the scene may be detected by the sensor system and appear as bright spots in the images generated for such scenes. When a light ray strikes a surface, the light can be reflected from the surface. Such a ray is also called an incident ray. Light can be reflected from the surface of various types of objects in the scene. Such objects include, for example, without limitation, people, buildings, aircraft, automobiles, leaves, artificial structures, and other suitable types of objects.

   The reflection of light on the surface is classified into diffuse reflection and specular reflection. Diffuse reflection occurs when incident light is reflected in multiple directions from the surface.

   Specular reflection occurs when incident light is reflected from the surface primarily in one direction. Specifically, specular reflection occurs when incident light and reflected light have the same angle with respect to an axis that is orthogonal to the surface. Specular reflections in the scene may appear as bright spots in the image generated by the sensor system. A bright spot is an area in an image where the intensity value of the pixels contained therein is greater than some threshold selected for luminance.

   When specular reflection is caused by light reflected on a large surface area, this reflection is called “glare”. When specular reflection is light reflected on a small surface area, this reflection is called “glint”. Glints may appear as small features in the image that have greater intensity compared to the surrounding image portions. In other words, glint may appear as a small bright spot in the image. Typically, bright spots in an image caused by glint are undesirable features in the image.

   When this type of undesirable feature occurs in an image, it becomes unnecessarily difficult to identify and / or analyze object features in the image of the scene. For example, when generating an image of an object, if undesired features are present in the image, identification of the object is more difficult than necessary or takes more time than necessary. Such undesirable features can make it more difficult to identify object features that are used to identify the object itself.

   As another example, when acquiring successive images of an object at different times, such images can be compared with each other to determine if there is a change in the object. The image can be used to determine if there is an inconsistency in subsequent images of the object. The presence of undesirable features in these images, such as bright spots caused by glints, makes it more difficult to identify inconsistencies and takes more time to identify.

   Some currently available systems for reducing glint in images use high speed optical machines attached to the imaging system. These optical machines include, for example, a polarizing plate and an optical filter. These types of optical machines can reduce the overall transmittance of light passing through the imaging system. By reducing the light transmittance, the amount of information supplied in the image generated by the imaging system can be reduced. In addition, these types of optical machines can unnecessarily increase the weight and cost of the imaging system.

   Therefore, it would be advantageous to have a method and apparatus that takes into account at least some of the problems discussed above and other possible problems.

   In one advantageous embodiment, a method for removing unwanted features in an image is provided. A transformed image is formed by transforming the image from the spatial domain to the frequency domain. Apply a filter to the transformed image to form a filtered image in the frequency domain. A modified image is formed by returning the filtered image from the frequency domain to the spatial domain. The intensity of undesired features in the modified image is greater than the intensity of undesired features in the original image. A processed image is formed by removing unwanted features from the image using the modified image.

   Another alternative embodiment considers using the processed image to identify features and / or objects in the image. In addition, the method uses the processed image to determine if there is a mismatch in the objects in the image. Also, using the processed image to determine whether there is a mismatch in the object further comprises forming a comparison spectrum by comparing the processed image with another image of the object, and Using the comparison spectrum to determine if a mismatch exists.

   In another advantageous embodiment, the apparatus comprises a computer system. The computer system forms a transformed image by transforming the image from the spatial domain to the frequency domain. The computer system further forms a filtered image in the frequency domain by applying a filter to the transformed image. The computer system further forms a modified image by transforming the filtered image back from the frequency domain to the spatial domain. The intensity of undesired features in the modified image is greater than the intensity of undesired features in the original image. The computer system further forms a processed image by removing unwanted features from the image using the modified image.

   The apparatus can also optionally include a sensor system that generates an image of the scene, where the undesirable features in the image are bright spots in the image caused by glint in the scene. Also considered is a device that forms a transformed image by transforming the image from the spatial domain to the frequency domain and computing the Fourier transform of the image that transforms the image from the spatial domain to the transformed image in the frequency domain. Is to include a computer system configured.

   In a configuration that forms a filtered image in the frequency domain by applying a filter to the transformed image, the device uses the root filter to form the filtered image to the phase of the phase component of the Fourier transform. It is also possible to have a computer system that scales the amplitude of the amplitude component of the Fourier transform of the image without affecting it. In this case, the root filter scales the amplitude nonlinearly.

   According to other optional mechanism considerations, in a configuration that uses the modified image to form a processed image by removing undesirable features from the image, the computer system applies a threshold to the modified image. Form a thresholded image by retaining pixels in the image having a value greater than the threshold; convolve the modified image with a point spread function of the image to form a convolved image, where The convolved image indicates the first location of the undesired feature in the convolved image, which corresponds to the second location of the undesired feature in the original image. And forming a processed image by subtracting the convolved image from the original image, wherein the image portion of the second location of the undesired feature in the original image By being removed from the original image, unwanted features are removed.

   In another advantageous embodiment, the vehicle inspection system comprises a sensor system, an image processor, and an image analyzer. The sensor system generates an image of the vehicle. The image processor receives an image from the sensor system. The image processor further forms a transformed image by transforming the image from the spatial domain to the frequency domain. The image processor further forms a filtered image in the frequency domain by applying a filter to the transformed image. The image processor further forms a modified image by returning the filtered image from the frequency domain to the spatial domain. The intensity of undesired features in the modified image is greater than the intensity of undesired features in the original image. The image processor further forms a processed image by removing unwanted features from the image using the modified image. The image analyzer analyzes the processed image to determine if there is a mismatch in the vehicle.

   Other vehicle inspection systems considered may include a display system, and the image analyzer may provide a display on the display system that indicates whether there is a mismatch in the vehicle. Another preferred image analyzer can further generate a report that indicates whether there is a mismatch in the vehicle.

   The above features, functions and advantages can be implemented independently in various embodiments of the invention or may be combined in other embodiments. These embodiments will be described in more detail with reference to the following description and attached drawings.

   Advantageous embodiment features that are considered novel features are set forth in the appended claims. However, advantageous embodiments, preferred modes of use, and further objects and advantages thereof are best understood by reading the following detailed description of one advantageous embodiment of the invention with reference to the accompanying drawings. It will be.

FIG. 6 is a block diagram of an image processing environment in accordance with an advantageous embodiment. FIG. 6 is a block diagram of an image processor according to an advantageous embodiment. Fig. 4 shows a panchromatic image according to an advantageous embodiment; Figure 8 illustrates a threshold application image according to an advantageous embodiment; Fig. 4 illustrates a processed image according to an advantageous embodiment. Fig. 4 shows an amplitude spectrum according to an advantageous embodiment; FIG. 5 is a flow diagram of a process for removing undesirable features from an image according to an advantageous embodiment. FIG. 6 is a flow diagram of an image analysis process according to an advantageous embodiment. FIG. 6 is a flow diagram of an image analysis process according to an advantageous embodiment. 1 illustrates a data processing system according to an advantageous embodiment;

   Various advantageous embodiments recognize and take into account one or more considerations. For example, the various advantageous embodiments recognize and take into account that images of a scene can be generated using different wavelengths. In other words, multiple wavelength measurements can be performed on objects in the scene. By generating images of the scene using different wavelengths, features that appear in the image in response to reflections (eg, glint) in the scene can be identified and removed from the image.

   However, the various advantageous embodiments recognize and take into account that some sensors cannot perform this type of wavelength measurement. For example, panchromatic image sensors do not provide this type of information. The panchromatic image sensor only provides a single band in which the image contains data in gray shades.

   Various advantageous embodiments also recognize and take into account that filtering can be performed to suppress undesirable features in the data related to the image. However, various advantageous embodiments can be more difficult and / or time consuming to differentiate unwanted features caused by glints from other features in an image of an object in a scene. Recognize and take this into account.

   In addition, the various advantageous embodiments recognize and take into account that currently available glint reduction systems using high speed optical machines cannot remove glint in previously generated images. For example, historical data used for testing and evaluating objects includes images that are weeks, months, or years old. Various advantageous embodiments recognize and take into account that it is desirable to reduce glint in these previously generated images.

   The various advantageous embodiments also recognize and take into account that currently available glint reduction systems cannot be adapted to changing conditions such as environmental conditions. For example, currently available imaging systems with polarizing plates that reduce glint in images generated by the imaging system may take into account changing weather conditions, wind conditions, and / or other types of environmental conditions. It is impossible to adjust the polarizing plate. As a result, some of the images generated by the imaging system may have an undesirable number of glints.

In addition, some currently available glint reduction systems can use motorized optical filters that adjust the polarization of the lenses of the imaging system. Various advantageous embodiments recognize and take into account that these types of systems cannot increase the number of frames generated per second in image processing. As a result, it is not possible with currently available systems to reduce glint in image processing in near real time.
Accordingly, various advantageous embodiments provide a method and apparatus for removing undesirable features caused by glints from an image. In various advantageous embodiments, undesired features are features that increase the difficulty of image analysis and / or increase the time taken for image analysis.

   In one advantageous embodiment, a method for removing unwanted features in an image is provided. A transformed image is formed by transforming the image from the spatial domain to the frequency domain. A filtered image is formed in the frequency domain by applying a filter to the transformed image. A modified image is formed by returning the filtered image from the frequency domain to the spatial domain. The intensity of undesired features in the modified image is greater than the intensity of undesired features in the original image. A processed image is formed by removing unwanted features from the image using the modified image.

   This information can then be used for any number of different purposes. For example, the processed image is used to identify objects in the scene captured in the image, identify inconsistencies in the object, identify object characteristics, and / or other objects related to objects in the image Appropriate operations can be performed.

   Reference is now made to FIG. FIG. 1 is a block diagram of an image processing environment in accordance with an advantageous embodiment. In such an embodiment, image processing environment 100 includes scene 102, sensor system 104, and computer system 105. The sensor system 104 generates an image 106 of the scene 102.

   As shown, sensor system 104 includes any number of sensors 108. As used herein, “any number of items” means one or more items. For example, “any number of sensors” means one or more computers. In such an embodiment, any number of sensors 108 may take the form of any number of imaging sensors 110 that are configured to detect electromagnetic radiation within band 112.

   The band 112 is a continuous band of frequency or wavelength. When the band 112 is in the visible spectrum of electromagnetic radiation, any number of imaging sensors 110 produce an image 106 of the sensor 102 in the form of a panchromatic image 114. When the band 112 is in another part of the spectrum of electromagnetic radiation, the image 106 is called a broadband image.

   In such an embodiment, image 106 takes the form of a panchromatic image 114. The panchromatic image 114 only includes a gray shade ranging from black to white. Specifically, the value of a pixel in such type of image contains only intensity information. In other words, the value of the pixel in the panchromatic image 114 indicates the intensity, not the color.

   As an example, any number of objects 116 exist in the scene 102. Any number of objects 116 in the scene 102 may include, but are not limited to, vehicles, aircraft, automobiles, people, buildings, man-made structures, roads, trees, green spaces, sea levels, lake surfaces, and other suitable types. Contains objects.

   Any number of imaging sensors 110 are configured to detect the brightness of any number of objects 116 in the scene 102 and generate an image 118. Image 118 is a two-dimensional image in such an embodiment. Specifically, any number of imaging sensors 110 measure the intensity of light in the band 112 detected by any number of sensors 110 and generate a value for each pixel in the image 118. For example, the value is about 0-255. Thus, the image 118 is a panchromatic image.

   The sensor system 104 sends the image 118 to the computer system 105 for processing. In such an embodiment, computer system 105 includes any number of computers. When there are a plurality of computers in the computer system 105, these computers can communicate with each other.

   As shown, there is an image processor 120 and an image analyzer 122 in the computer system 105. Image processor 120 and image analyzer 122 may be implemented in computer system 105 using hardware, software, or a combination of the two. In some embodiments, these modules can be used independently of other components in computer system 105.

   These modules, when implemented as hardware, can be implemented using any number of circuits configured to perform the desired functions and / or processes. Any number of such circuits include, for example, integrated circuits, application specific integrated circuits, programmable logic arrays, general purpose logic arrays, field programmable gate arrays, programmable logic devices, complex programmable logic devices (CPLDs), programmable logic controllers. , Macro cell arrays, and other suitable types of circuits.

   As used herein, the expression “at least one of” used with the listed items can be used in various combinations of one or more of the listed items. This means that only one item is required. For example, “at least one of item A, item B, and item C” includes, for example, without limitation, “item A” or “item A and item B”. This example also includes “item A and item B and item C” or “item B and item C”. As another example, “at least one of” is, for example, without limitation, “two items A, one item B, and ten items C”, “four items B and 7”. Item C ”as well as other suitable combinations.

   Further, in such an embodiment, computer system 105 also includes a display system 124. Display system 124 may include one or more display devices. In such examples, the various components within computer system 105 may be in the same location or different locations depending on the particular implementation.

   In some embodiments, computer system 105 may be a distributed computer system, such as a network data processing system. In other embodiments, the computer system 105 may be entirely composed of hardware circuits in which processes are implemented in hardware rather than software.

   In such an embodiment, image processor 120 receives and processes image 118. Specifically, the image processor 120 processes the image 118 to remove unwanted features 126 from the image 118. More particularly, the image processor 120 processes the image 118 to remove unwanted features 126 from the image 118 caused by unwanted reflections 130 in the scene 102. Specifically, the unwanted reflection 130 is a glint 131.

   The glint 131 is a specular reflection of light from the surface of any number of objects 116 in the scene 102 that appear in the image 118. The glint appears in the image 118 as a small bright spot that has a greater intensity than the portion of the image 118 that surrounds it. In this way, undesirable features 126 caused by glint 131 in scene 102 take the form of undesirable bright spots 128 in image 118. The image processor 120 cannot use the current form of the image 118 to identify which bright spots in the image 118 are caused by the glint 131.

   For example, the image processor 120 can detect bright spots 132 in the image 118. The bright spot 132 may include one pixel, two pixels, or some other suitable number of pixels. However, the image processor 120 cannot determine whether the bright spot 132 is caused by a glint 134 in the scene 102 and is one of the unwanted bright spots 128 in the image 118. For example, the intensity of the bright spot 132 in the image 118 is not high enough to determine whether the bright spot 132 is caused by a glint 134 in the scene 102. As a result, the image processor 120 cannot determine whether the bright spot 132 is one of the unwanted bright spots 128 or a feature of any number of objects 116 in the scene 102.

   In such an embodiment, the current form of image 118 is in spatial domain 136. Spatial domain 136 is the image space of image 118. Specifically, the spatial domain 136 is a two-dimensional space for the image 118. In one embodiment, the spatial domain 136 is defined using the xy coordinate system of the image 118. The x coordinate of the image 118 represents the horizontal direction of the image 118, and the y coordinate represents the vertical direction of the image 118.

   Image processor 120 forms transformed image 140 by transforming image 118 from spatial domain 136 to frequency domain 138. Specifically, the image processor 120 computes the Fourier transform (FT) 142 of the image 118 and transforms the image 118 from the spatial domain 136 to the frequency domain 138 to form the transformed image 140. In such an embodiment, the Fourier transform 142 is either one selected from a discrete Fourier transform (DFT) and a fast Fourier transform (FFT). When the spatial domain 136 is a two-dimensional space, the Fourier transform 142 is also a two-dimensional Fourier transform.

   In such an embodiment, transformed image 140 includes an amplitude component 144 and a phase component 146. The amplitude component 144 specifies the amplitude of the Fourier transform 142 of the image 118. Such amplitude is also called magnitude. Phase component 146 identifies the phase of Fourier transform 142 of image 118.

   The amplitude component 144 is represented by an amplitude spectrum, for example. The phase component 146 is represented by a phase spectrum, for example. The amplitude spectrum and the phase spectrum indicate the amplitude and phase of the Fourier transform 142 in terms of frequency, respectively.

   In such an embodiment, transformed image 140 has the same number of pixels as image 118. Specifically, the converted image 140 has the same pixel dimensions as the image 118.

   Each pixel in the transformed image 140 represents a different spatial frequency 148 in the image 118. Spatial frequency 148 in image 118 is the rate of change of the value of a pixel relative to neighboring pixels in image 118. Specifically, spatial frequency 148 is defined as the number of changes that have occurred in the value of pixel intensity per selected distance 150 for a particular portion of image 118. The selected distance 150 may be in units of the transformed image 140, such as pixels, millimeters, centimeters, or some other suitable distance unit.

   As described above, the change in position in the converted image corresponds to the change in the spatial frequency 148. Further, the value of the pixel in the transformed image 140 indicates the difference between the brightest and darkest gray level within the selected distance 150 at the particular spatial frequency 148 that the pixel represents. Such a change from the brightest gray level to the darkest gray level within the selected distance 150 is referred to as a section or period.

   For example, if a pixel in the transformed image 140 representing a period of spatial frequency 148 for every 10 pixels has a value of 20, the pixel intensity value in the image 118 for 10 pixels will be the darkest to the darkest gray level. Change to a gray level or change from the darkest level to the brightest gray level. Furthermore, the contrast between the brightest and darkest gray levels is approximately twice the pixel value. In other words, such contrast is about 40 gray levels between the brightest and darkest gray levels.

   In such an embodiment, small features in image 118 have a higher spatial frequency compared to larger features. As a result, small features in the image 118 are represented in the transformed image 140 with higher frequency components compared to larger features in the image 118.

   In other words, such small features in the image 118 appear in pixels corresponding to high spatial frequencies in the transformed image 140 when compared to features larger than that in the image 118 and are large in the image 118. Features appear in pixels that correspond to low spatial frequencies in the transformed image 140. As an example, undesirable bright spots 128 in image 118 caused by glint 131 may be small features in image 118 and appear in high frequency components in transformed image 140.

   In such an embodiment, image processor 120 forms filtered image 154 in frequency domain 138 by applying filter 152 to transformed image 140. The filter 152 may take the form of, for example, without limitation, a non-linear filter, a linear filter, a high frequency pass filter, and / or any other suitable type of filter.

   In such an embodiment, the filter 152 is configured to change the amplitude component 144 of the Fourier transform 142 of the image 118 without changing the phase component 146. Specifically, the filter 152 changes the amplitude of the high frequency component of the converted image 140.

   More specifically, when the filter 152 is applied to the transformed image 140, the amplitude of the high frequency component in the transformed image 140 increases relative to the amplitude of the low frequency component in the transformed image 140. In other words, the ratio between the amplitude of the high frequency component and the amplitude of the low frequency component in the filtered image 154 is lower than the ratio of the amplitude of the high frequency component and the amplitude of the low frequency component in the converted image 140.

   In other words, since the amplitude of the high frequency components in the transformed image 140 is enhanced in the filtered image 154, the features in the image 118 corresponding to such high frequency components are also enhanced. Specifically, the amplitude of the high frequency component corresponding to the undesired bright spot 128 caused by the glint 131 in the image 118 is enhanced in the filtered image 154.

   The image processor 120 then returns the transformed image 140 in the frequency domain 138 to the spatial domain 136 by computing the inverse Fourier transform 155 of the transformed image 140. By such conversion, a corrected image 156 is formed. In particular, the intensity of undesirable bright spots 128 caused by small features in the modified image 156, such as glint 131, is greater than the intensity of such features in image 118. In this way, it is much easier in the modified image 156 than in the image 118 to distinguish unwanted bright spots 128 caused by glint 131 from other features.

   Image processor 120 then uses modified image 156 to remove undesirable bright spots 128 caused by glint 131 from modified image 156. Specifically, the image processor 120 identifies the location 158 of the unwanted bright spot 128 in the modified image 156 and then reduces and / or eliminates the unwanted bright spot 128 at the location 158 in the image 118. To do. In such an embodiment, processed image 160 is formed by reducing and / or removing unwanted bright spots 128 from modified image 156. In this way, undesirable bright spots 128 caused by glint 131 are suppressed.

   Image processor 120 sends processed image 160 to image analyzer 122 for analysis. The image analyzer 122 can analyze the processed image 160 in any number of different ways. For example, the image analyzer 122 can use the processed image 160 to identify any number of features 162 of the object 116. Features 162 of any number of objects 116 may include, for example, internal misalignment, edge misalignment, dimensional misalignment, identification misalignment, and / or identification of one object of any number of objects 116. Or any other suitable type of feature mismatch.

   In some embodiments, image processor 120 and image analyzer 122 are implemented in vehicle inspection system 164. Vehicle inspection system 164 is configured to use processed image 160 to determine whether inconsistency 166 exists in object 168 of any number of objects 116. In such an embodiment, object 168 is a vehicle. In such an embodiment, the mismatch in object 168 is a feature or feature in object 168 that should not exist or is undesirable in object 168.

   Identifying inconsistencies 166 can be performed in any number of different ways. For example, processed image 160 can be compared to image 170 stored in database 172. Image 170 is another image of object 168 that was generated prior to generation of image 118 of object 168. Such a comparison of the processed image 160 with the image 170 can be used to determine whether an inconsistency 166 exists.

   In other embodiments, the image processor 120 and the image analyzer 122 are implemented in the object identification system 174. In this type of implementation, the image analyzer 122 can generate an identification 176 of the object 168 by comparing the image 170 and the processed image 160. In some cases, the object 168 can be identified by comparing the processed image 160 with multiple images present in the database 172.

   In such an embodiment, vehicle inspection system 164 and / or object identification system 174 may use the analysis results of processed image 160 performed by image analyzer 122 to generate a display on display system 124. it can. In some embodiments, the image analyzer 122 can provide a display of the mismatch 166 in the object 168 on the display system 124.

   In other embodiments, the image analyzer 122 includes an indication of whether there is a mismatch 166 in the object 168, an identification 176 of the feature 162 of the object 168, an identification of the object 168, and / or other suitable information. Generate a report. This report may be displayed on the display system 124, emailed to the client, stored in a database, and / or managed in some other suitable manner, for example.

   Further, depending on the implementation, the image processor 120 and / or the image analyzer 122 are configured to display an image on the display system 124. For example, the image processor 120 and / or the image analyzer 122 can display either the image 118 and / or the processed image 160 on the display system 124.

   The image processing environment 100 shown in FIG. 1 is not intended to suggest physical or architectural limitations to the manner in which advantageous embodiments may be implemented. Other components can be used in addition to and / or instead of the components shown. Some components may be unnecessary. Blocks are also presented to show some functional components. When implemented in an advantageous embodiment, one or more of these blocks can be combined and / or divided into different blocks.

   For example, in some embodiments, the image analyzer 122 and the image processor 120 may be located at different locations. For example, the image processor 120 may be located on the same platform as the sensor system 104 and the image analyzer 122 may be located on some other platform remote from the sensor system 104. In other embodiments, the image processor 120 and the image analyzer 122 can be implemented in the same module in the computer system 105.

   Further, although described as a vehicle, the object 168 can take any number of different forms. For example, the object 168 may be a movable platform, a fixed platform, a land base structure, a water base structure, a space base structure, an aircraft, a watership, a tank, a personnel carrier, a train, a spacecraft, a space station, a satellite, a submarine, an automobile, a power generation One selected from a station, a bridge, a dam, a manufacturing plant, a building, a skin panel, an engine, a section of a gas piping system, and other suitable objects.

   Reference is now made to FIG. FIG. 2 is a block diagram of an image processor according to an advantageous embodiment. In this embodiment, the image processor 200 is an example of one implementation of the image processor 120 of FIG. As shown, the image processor 200 includes a transform unit 202, a filter unit 204, an inverse transform unit 206, a threshold unit 208, a point spread function estimator 210, a convolution unit 212, and a feature removal unit 214.

   In this example, conversion unit 202 receives image 216 from a sensor system (eg, sensor system 104 of FIG. 1). In this example, image 216 takes the form of panchromatic image 218. Further, the image 216 is in the spatial domain 200. The spatial domain 220 is a two-dimensional image space of the image 216. In this example, the image 216 in the spatial domain 220 is represented by a function, H (x, y).

   A function H (x, y) of the image 216 is a function for obtaining the value of the pixel with respect to the x, y position of the pixel in the image 216. The pixels in image 216 are organized in rows and columns. In such an embodiment, the x position of the pixel identifies a particular row and the y position identifies a particular column.

   A bright spot 221 appears in the image 216. The bright spot 221 is a small feature in the image 216 that has a greater intensity than the portion of the image 216 surrounding the bright spot 221. At least some of the bright spots 221 do not have enough intensity to distinguish bright spots caused by glint in the scene captured by the image 216 from bright spots that are features of objects in the scene. Sometimes. The image processor 200 processes the image 216 to make such a distinction.

   As shown, transform unit 202 forms transformed image 224 by transforming image 216 from spatial domain 220 to frequency domain 222. Specifically, the transform unit 202 forms a transformed image 224 by computing a Fourier transform (FT) 226 of the image 216. The Fourier transform 226 of the image 216 is represented by a function, H (u, v). A function of the image 216, H (u, v), is obtained by converting the function of the image 216, H (x, y), into the frequency domain 222.

   In such an embodiment, filter unit 204 receives transformed image 224 in frequency domain 222 from transform unit 202. The filter unit 204 generates a filtered image 230 by applying a route filter 228 to the converted image 224. In this embodiment, the root filter 228 is a non-linear filter that uses the parameters 232 and α to change the amplitude component of the Fourier transform 226 of the image 216 without changing the phase component of the Fourier transform 226.

Specifically, the parameters 232 and α are indices having values of 0 to 1. The root filter 228 increases the amplitude component of the Fourier transform 226 to the power of the parameter 232. This operation corresponds to taking the α th route of the amplitude component of the Fourier transform 226. In other words, the filtered image 230 is represented by the following equation:
F (u, v) = | H (u, v) | ^a e ^{jθ (u, v)} (1)
Where F (u, v) represents the filtered image 230, | H (u, v) | is the amplitude component of the Fourier transform 226 of the image 216, and u is the horizontal frequency in the frequency domain 222. Where v is the vertical component of the frequency in the frequency domain 222, α is a parameter 232 between 0 and 1, e is an exponential function, θ is the phase angle, and j is the square root of −1. is there.

   The value α of the parameter 232 is selected so that the amplitude of the high frequency component of the transformed image 224 is increased above the amplitude of the low frequency component. Specifically, the amplitude of the high frequency component of the converted image 224 is increased with a larger width than the amplitude of the low frequency component of the converted image 224. In some cases, the amplitude of the high frequency component is reduced within the transformed image 224 by a smaller width than the amplitude of the low frequency component.

   The filtered image 230 generated by the filter unit 204 is sent to the inverse transform unit 206. The inverse transform unit 206 operates the inverse Fourier transform 234 of the filtered image 230 in the frequency domain 222 to return the filtered image 230 from the frequency domain 222 back to the spatial domain 220 in the form of a modified image 236.

   In this example, bright spots 238 appear in the corrected image 236. The bright spot 238 is substantially the same as the bright spot 221 in the image 216. The bright spot 238 in the modified image 236 in the spatial domain 220 has a greater intensity than the bright spot 221 in the image 216. Specifically, the bright spot 238 has such an intensity that the bright spot caused by the glint in the scene captured by the image 216 can be distinguished from the bright spot that is a feature of the object in the scene.

As shown, threshold unit 208 receives modified image 236 from inverse transform unit 206. The threshold unit 208 applies the threshold values 240 and K _T to the corrected image 236. Specifically, when threshold 240 is applied to modified image 236, pixels in modified image 236 that have a value less than or equal to threshold 240, K _T are set to have approximately zero or some minimum amplitude. Pixels in the modified image 236 that have a value above the threshold 240, _KT are left intact, and a threshold application image 242 is formed in the spatial domain 220.

The threshold 240 is selected so that all bright spots 238 caused by glints in the scene captured by the image 216 can be identified. For example, the pixels in the modified image 236 that have a value above the threshold 240, _KT are the pixels of the set of bright spots 244 of the bright spots 238. As used herein, “a set of items” means zero or more items. For example, a set of items can possibly be a null value or an empty set. Specifically, the set of bright spots 244 includes the bright spots of the bright spots 238 that can be generated by the bright spots in the scene captured by the image 216. In some embodiments, the set of bright spots 244 is also referred to as a set of glint points.

   The set of bright spots 244 is located at a set of positions 245 in the threshold application image 242. In this example, parameter 232 is selected such that one of the set of bright spots 244 has a width of about 1 pixel. Specifically, the parameter 232 is selected such that the smallest bright spot in the set of bright spots 244 occupies about 1 pixel or less. Further, in some embodiments, the threshold unit 208 can be configured to apply more than one threshold to the modified image 236.

   The convolution unit 212 is configured to generate a convolved image 248 in the spatial domain 220 by convolving the threshold application image 242 with the point spread function 246 of the image 216. In this example, point spread function 246 is generated by point spread function estimator 210. Point spread function 246 represents the response of the sensor system that generated image 216 to a point source or point object.

   For example, a point feature in the scene 102 of FIG. 1 is a feature of an object in the scene that occupies approximately one pixel in the image 216 based on, for example, the distance between the sensor system and the object. However, in a sensor system that produces image 216, this point feature may be blurred in image 216, resulting in a point feature occupying more than one pixel. The point spread function 246 represents such blurring of point features. In other words, the point spread function 246 represents the blurring of point features by the system that generated the image 216. In such an embodiment, the point spread function 246 can take the form of a Gaussian function, p (x, y).

   In one embodiment, the point spread function estimator 210 uses the image 216 to estimate the point spread function 246. In other embodiments, the point spread function estimator 210 may obtain the point spread function 246 from user input, a database, or some other suitable source. For example, the parameters of the sensor system that generated the image 216 are well known, and the sensor system point spread function 246 has already been defined.

   In such an embodiment, the convolution unit 212 generates a convolved image 248 by convolving the point spread function 246 with the threshold application image 242, which produces a set of locations 245 of a set of bright spots 244. To form a final location set 252 for the final bright spot set 250. The final bright spot set 250 located in the final position set 252 estimates the glint contribution to the image 216.

   Specifically, the final set of positions 252 takes into account the effects of blurring due to the sensor system. In other words, the final set of positions 252 is a position corresponding to the position of the corresponding bright spot among the bright spots 221 in the image 216. The convolution of the estimated point spread function 246 and the threshold application image 242 function in the spatial domain 220 is to integrate the product of these two functions after inverting one and shifting.

   As shown, feature removal unit 214 is configured to receive convolved image 248. In some embodiments, feature removal unit 214 applies gain 253 to convolved image 248 before receiving convolved image 248. In such an embodiment, gain 253 is selected to account for the type of sensor system that generated image 216, the glint in the scene that may affect the bright spots in image 216, and / or other suitable factors. Is done.

   Feature removal unit 214 subtracts convolved image 248 from image 216. In other words, the portion of image 216 located at a position in image 216 corresponding to final position set 252 of final bright spot set 250 in convolved image 248 is removed from image 216 and A processed image 256 is formed. Such removal is removal of unwanted features 254 from image 216. Specifically, such removal is removal of undesirable features 254 caused by glint in the scene captured by image 216.

   In this example, the removal of undesired features 254 may result in the pixel values of the portion of the image 216 located at a location corresponding to the final location set 252 of the final bright spot set 250 being zero or non-zero. This is done by setting it to some minimum value. A minimum value other than zero is selected to take into account the possibility that one of the bright spots 221 in the image 216 is caused by a reflection other than glint. For example, bright spots can also occur due to diffuse reflection.

   Image processor 200 sends processed image 256 to an image analyzer (eg, image analyzer 122 of FIG. 1). In this way, analysis can be performed on the processed image 256 from which the undesired features 254 have been removed, rather than the image 216 in which the undesired features 254 still exist. Analysis of processed image 256 is more accurate than analysis of image 216.

   The image processor 200 shown in FIG. 2 is not intended to suggest physical or architectural limitations to the manner in which advantageous embodiments may be implemented. In such embodiments, image processor 200 may be implemented using other types of components in addition to and / or instead of the illustrated components. In addition, some of the illustrated components can be combined with other components, or the functionality can be further separated into additional components.

   For example, the threshold unit 208 can be implemented as part of the feature removal unit 214. In some embodiments, a bidirectional conversion unit may be provided that performs the functions performed by both the conversion unit 202 and the inverse conversion unit 206.

   FIG. 3 shows a panchromatic image according to an advantageous embodiment. The panchromatic image 300 is an example of one of the image 106 in FIG. 1 and the panchromatic image 218 in FIG. The panchromatic image 300 is an image of a section of the aircraft fuselage. The panchromatic image 300 is a gray scale image. As shown, a bright spot 302 exists in the panchromatic image 300. At least a part of the bright spot 302 is caused by a glint in the scene captured by the panchromatic image 300. A bright spot 304 is an example of a bright spot 302 that is at least partially caused by glint. The bright spot 304 is located at a position 306 in the panchromatic image 300.

   FIG. 4 shows a threshold application image according to an advantageous embodiment. The threshold application image 400 is an example of one implementation of the threshold application image 242 of FIG. Specifically, the threshold application image 400 is an example of an image generated by the threshold unit 208 of FIG. 2 after the panchromatic image 300 of FIG. 3 is processed by the image processor 200 of FIG.

   As shown, the threshold application image 400 is a black and white image that includes pixels 402 having zero values and pixels 404 having non-zero values. A pixel 404 having a non-zero value identifies a position in the threshold application image 400 that corresponds to the position of the bright spot 302 portion in the panchromatic image 300 of FIG.

   FIG. 5 illustrates a processed image according to an advantageous embodiment. The processed image 500 is an example of one implementation of the processed image 256 of FIG. Specifically, processed image 500 is an example of an image generated by image processor 200 of FIG. 2 after processing panchromatic image 300 of FIG. 3 to remove unwanted features from panchromatic image 300. Such undesirable features are bright spot 302 portions in the panchromatic image 300 of FIG. 3 caused by glint.

   Specifically, a convolved image is formed by convolving the threshold application image 400 of FIG. 4 using a point spread function of the panchromatic image 300. Such convolution is performed by a convolution unit 212 in the image processor 200 of FIG. The feature removal unit 214 in the image processor 200 of FIG. 2 generates a processed image 500 by subtracting the convolved image from the panchromatic image 300. As shown, the processed image 500 is a grayscale image.

   In such an embodiment, position 502 in processed image 500 corresponds to position 306 in panchromatic image 300. As shown in the processed image 500, the glint effects from the fuselage section of the aircraft in the processed image 500 have been largely eliminated at location 502. Specifically, when compared with the position 306 in the panchromatic image 300, the influence of glint at the position 502 is substantially eliminated. The luminance remaining in location 502 is the effect of diffuse reflection from the aircraft fuselage compartment.

   Reference is now made to FIG. FIG. 6 shows an amplitude spectrum according to an advantageous embodiment. In this example, plot 600 has a horizontal axis 602 and a vertical axis 604. The horizontal axis 602 is a spatial frequency indicated by a cycle per sample. The vertical axis 604 is the amplitude. As shown, plot 600 includes an amplitude spectrum 606 and an amplitude spectrum 608.

   The amplitude spectrum 606 is an example of the expression of the amplitude component of the converted image 224 (for example, the converted image 224 in FIG. 2). The amplitude spectrum 608 is an example of a representation of the amplitude component of the filtered image (eg, filtered image 230 of FIG. 2). The filtered image is a result of applying the route filter 228 of FIG. 2 to the converted image 224.

   As illustrated, the amplitude spectrum 606 and the amplitude spectrum 608 indicate that the ratio of the high frequency component amplitude to the low frequency component amplitude in the filtered image 230 is the ratio of the high frequency component amplitude to the low frequency component amplitude in the transformed image 224. It is shown to be smaller. In other words, when the root filter 228 is applied to the converted image 224, the amplitude of the high frequency component increases compared to the amplitude of the low frequency component. In this example, when the root filter 228 is applied, the amplitude of all frequencies of the transformed image 224 is reduced.

   Reference is now made to FIG. FIG. 7 is a flow diagram of a process for removing unwanted features from an image according to an advantageous embodiment. The process illustrated in FIG. 7 may be implemented in the image processing environment 100 of FIG. In particular, one or more of the various advantageous embodiments can be implemented in the image processor 120 of FIG. In addition, depending on the particular implementation, several steps may be performed using the image analyzer 122 of FIG.

   The process begins by receiving an image of the scene generated by the sensor system (step 700). In this embodiment, the sensor system includes any number of panchromatic image sensors that generate images as panchromatic images. In this embodiment, the scene image is generated in the spatial domain.

   Next, the process forms a transformed image by transforming the image from the spatial domain to the frequency domain (step 702). Specifically, in step 702, the process computes a Fourier transform of the image that transforms the image in the spatial domain into a transformed image in the frequency domain. The Fourier transform of an image includes an amplitude component and a phase component. Further, in this embodiment, the Fourier transform can be a discrete Fourier transform (DFT) or a fast Fourier transform (FFT).

   The process then forms a filtered image in the frequency domain by applying a filter to the transformed image (step 704). In this example, the filter in step 704 is a root filter. Specifically, the root filter scales the amplitude of the amplitude component of the Fourier transform without changing the phase of the phase component of the Fourier transform. The root filter is configured to scale the amplitude of the amplitude component of the Fourier transform nonlinearly.

   The process then forms a modified image by returning the filtered image from the frequency domain to the spatial domain (step 706). In step 706, the filtered image in the frequency domain is converted to a modified image in the spatial domain by computing an inverse Fourier transform of the filtered image. The intensity of all undesired features in the modified image caused by glint in the scene is increased over the intensity of undesired features in the image of the scene. In other words, the intensity of all undesired features in the modified image is enhanced relative to other parts of the image when compared to the intensity of undesired features in the image generated by the sensor system.

   The process then forms a threshold application image by applying a threshold to the modified image (step 708). In step 708, pixels in the modified image having values above the threshold are retained to form a threshold application image. All pixels that are retained contain unwanted features caused by glint in the scene detected by the sensor system. Specifically, the corrected image includes a set of bright spots. This set of bright spots corresponds to small features that appear in the image generated by the sensor system due to glint in the scene.

   Next, the process forms a convolved image by convolving the modified image using a point spread function of the image (step 710). The convolved image shows a set of positions corresponding to the set of bright spots identified in the modified image. The set of positions takes into account the effects of blur generated by the sensor system.

   The process then removes all undesired features caused by the glint to form a processed image by subtracting the convolved image from the image generated by the sensor system (step 712) and then ends. . In step 712, the portion of the image located at a position corresponding to a set of bright spots in the corrected image is subtracted from the image generated by the sensor system. Everything is removed.

   Reference is now made to FIG. FIG. 8 is a flow diagram of an image analysis process according to an advantageous embodiment. The process illustrated in FIG. 8 may be implemented in the image processing environment 100 of FIG. Specifically, this process is performed by the image analyzer 122 of FIG.

   The process begins by receiving a processed image (step 800). This processed image is an image generated by the image processor 120 of FIG. 1 using the process described in FIG. Specifically, this processed image can be the processed image formed in step 712 of FIG.

   The process analyzes the processed image and generates a result (step 802). It is determined whether the results indicate the presence of inconsistencies in the objects in the image (step 804). If an inconsistency exists in the object, the process generates an indication of the presence of the inconsistency (step 806) and then ends. This display may be provided on a display device, emailed, stored in a database, or processed in some other suitable manner.

   Returning to step 804, if there is no inconsistency in the object, the process generates an indication of the presence of this inconsistency (step 808) and then ends.

   The analysis performed in step 802 can be performed in any number of different ways. For example, the analysis can be performed directly on the processed image. The data in the processed image may be compared with other images of the previously imaged object. Such a comparison can be used to determine whether an object has changed. Such a change can be used to determine if an inconsistency exists in the object. In such an embodiment, the inconsistency can be an undesirable change in the object.

   Reference is now made to FIG. FIG. 9 is a flow diagram of an image analysis process according to an advantageous embodiment. The process shown in FIG. 9 can be implemented in the image processing environment 100 of FIG. Specifically, this process is performed by the image analyzer 122 of FIG.

   The process begins by receiving a processed image (step 900). This processed image is an image generated by the image processor 120 using the process described in FIG. Specifically, this processed image can be the processed image formed in step 712 of FIG.

   The process then identifies the objects in the image by analyzing the processed image (step 902). This analysis can be performed in any number of different ways. For example, a database of previous images or other data can be used as a comparison with the data in the processed image. Based on such comparison, the object can be identified. This identification suggests that a particular object is identified or that the object in the image is unknown.

   The process then generates an indication of the object's identity (step 904) and then ends. This display is, for example, a report for identifying an object.

   Reference is now made to FIG. FIG. 10 illustrates a data processing system according to an advantageous embodiment. In such an embodiment, data processing system 1000 may be used to implement one or more computers included in computer system 105 of FIG. In this illustrative example, data processing system 1000 includes a communication framework 1002 that enables processor unit 1004, memory 1006, persistent storage 1008, communication unit 1010, input / output (I / O) unit 1012, and display. Communication between the units 1014 is performed.

   The processor unit 1004 serves to execute instructions for software loaded into the memory 1006. The processor unit 1004 may be any number of processors, multiprocessor cores, or some other type of processor, depending on the particular implementation. References herein to an item and “an arbitrary number” refer to one or more items. Further, the processor unit 1004 may be implemented using any number of heterogeneous processor systems in which a main processor and a secondary processor coexist on a single chip. In another embodiment, the processor unit 1004 may be a symmetric multiprocessor system including a plurality of processors of the same type.

   The memory 1006 and the fixed storage area 1008 are examples of the storage device 1016. A storage device is any piece of hardware that can store information temporarily and / or permanently, such as, but not limited to, data, functional form program code, and / or other suitable Information. The storage device 1016 is also referred to as a computer-readable storage device in such embodiments. In such an embodiment, memory 1006 may be, for example, random access memory or some other suitable volatile or non-volatile storage device. The persistent storage 1008 can take a variety of forms depending on the particular implementation.

   For example, persistent storage 1008 may include one or more components or devices. For example, persistent storage 1008 is a hard drive, flash memory, rewritable optical disk, rewritable magnetic tape, or some combination thereof. The medium used by persistent storage 1008 may be removable. For example, a removable hard drive can be used for persistent storage 1008.

   In such an embodiment, the communication unit 1010 communicates with other data processing systems or devices. In such an embodiment, the communication unit 1010 is a network interface card. The communication unit 1010 can perform communication by using one or both of a physical communication link and a wireless communication link.

   The input / output unit 1012 allows data input and output by other devices that can be connected to the data processing system 1000. For example, input / output unit 1012 may provide a connection for user input via a keyboard, mouse, and / or any other suitable input device. Further, the input / output unit 1012 can send output to a printer. Display 1014 provides a mechanism for displaying information to the user.

   Instructions for the operating system, applications, and / or programs may be located in the storage device 1016 that communicates with the processor unit 1004 via the communication framework 1002. In such an embodiment, the instructions are in functional form on persistent storage 1008. These instructions are loaded into the memory 1006 and executed by the processor unit 1004. The processes of various embodiments may be performed by processor unit 1004 using computer-implemented instructions located in memory (eg, memory 1006).

   These instructions are called program code, computer usable program code, or computer readable program code, and may be read and executed by one processor in processor unit 1004. The program code of the various embodiments may be embodied on various physical storage media or computer readable storage media (eg, memory 1006 or persistent storage 1008).

   Program code 1018 is placed in a functional form on a selectively removable computer readable medium 1020, loaded or transferred to data processing system 1000 and executed by processor unit 1004. Program code 1018 and computer readable media 1020 form computer program product 1022 in these examples. In one embodiment, computer readable medium 1020 is a computer readable storage medium 1024 or a computer readable signal medium 1026.

   The computer-readable storage medium 1024 is inserted or arranged in a drive or other device that is a part of the fixed storage area 1008, for example, on a storage device such as a hard disk that is a part of the fixed storage area 1008. It may include an optical disk or a magnetic disk to be transferred. The computer readable storage medium 1024 may take the form of a fixed storage area such as a hard drive, a thumb drive, or a flash memory connected to the data processing system 1000. In some examples, computer readable storage media 1024 may not be removable from data processing system 1000.

   In such embodiments, computer readable storage medium 1024 is a physical or tangible storage device that is used to store program code 1018 rather than a medium that propagates or transfers program code 1018. The computer-readable storage medium 1024 is also called a tangible storage device that can be read by a computer or a physical storage device that can be read by a computer. In other words, the computer-readable storage medium 1024 is a medium that can be touched by a person.

   Alternatively, program code 1018 can be transferred to data processing system 1000 using a computer readable signal medium 1026. The computer readable signal medium 1026 may be, for example, a propagated data signal that includes program code 1018. For example, computer readable signal media 1026 may be an electromagnetic signal, an optical signal, and / or any other suitable type of signal. These signals are transferred over a communication link such as a wireless communication link, fiber optic cable, coaxial cable, wired, and / or any other suitable type of communication link. In other words, in an embodiment of the present invention, the communication link and / or connection may be physical or wireless.

   In some advantageous embodiments, the program code 1018 is downloaded from another device or data processing system over a network to the persistent storage 1008 via a computer readable signal medium 1026 and stored in the data processing system 1000. used. For example, program code stored in a storage medium readable by a computer of the server data processing system can be downloaded from the server to the data processing system 1000 via a network. The data processing system providing program code 1018 may be a server computer, a client computer, or some other device capable of storing and transferring gram code 1018.

   The various components illustrated in data processing system 1000 are not architecturally limited in how various embodiments can be implemented. Various advantageous embodiments may be implemented in a data processing system that includes components in addition to or in place of those illustrated in data processing system 1000. Other components shown in FIG. 10 can be modified from the illustrated embodiment. Various embodiments may be implemented using any hardware device or system capable of executing program code. As an example, the data processing system can include an organic component integrated with an inorganic component and / or can consist entirely of organic components excluding humans. For example, the storage device can be formed of an organic semiconductor.

   In another embodiment, processor unit 1004 may take the form of a hardware unit having circuitry that is manufactured or configured for a particular application. This type of hardware can perform a process without having to load program code into memory from a storage device configured to perform the process.

   For example, if the processor unit 1004 takes the form of a hardware unit, the processor unit 1004 may be a circuit system, an application specific integrated circuit (ASIC), a programmable logic device, or some other suitable performing any number of steps. It may be a kind of hardware. In the case of a programmable logic device, the device performs any number of steps. The device can then be reconfigured or can be permanently configured to perform any number of steps. Examples of programmable logic devices include, for example, programmable logic arrays, field programmable logic arrays, field programmable gate arrays, and other suitable hardware devices. In this type of implementation, the program code 1018 may be omitted because the processes of the various embodiments are implemented on a hardware device.

   In yet another embodiment, processor unit 1004 may be implemented utilizing a combination of processors found in computers and hardware units. The processor unit 1004 may have any number of hardware units and any number of processors configured to execute the program code 1018. In this example, some of the processes can be performed on any number of hardware units, and other processes can be performed on any number of processors.

   In another embodiment, the bus system can be used to implement the communication framework 1002 and can consist of one or more buses, such as a system bus or an input / output bus. Of course, the bus system may be implemented using any suitable type of architecture that provides data transmission between various components or devices attached to the bus system.

   In addition, a communication unit may include any number of devices that transmit data, receive data, or transmit and receive data. The communication unit may be, for example, a modem or network adapter, two network adapters, or some combination thereof. Further, the memory may be, for example, memory 1006 or a cache, such as found in interfaces and memory controller hubs that may exist within the communication framework 1002.

   Accordingly, various advantageous embodiments provide a method and apparatus for removing undesirable features caused by glints from an image. In various advantageous embodiments, undesired features are features that increase the difficulty of image analysis and / or increase the time taken for image analysis.

   In one advantageous embodiment, a method for removing unwanted features in an image is provided. A transformed image is formed by transforming the image from the spatial domain to the frequency domain. A filtered image is formed in the frequency domain by applying a filter to the transformed image. A modified image is formed by returning the filtered image from the frequency domain to the spatial domain. The intensity of undesired features in the modified image is greater than the intensity of undesired features in the original image. By using the modified image, unwanted features are removed from the image to form a processed image.

The descriptions of the various advantageous embodiments described above are for purposes of illustration and description, and are not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to practitioners skilled in this art. Furthermore, the various advantageous embodiments may provide different advantages than the other advantageous embodiments. The selected embodiment (s) are intended to best explain the principles of the embodiments, practical applications, and to others skilled in the art in terms of the disclosure of the various embodiments and the specific applications considered. Selected and described to facilitate understanding of various suitable modifications.
Moreover, this invention includes the aspect described below.
(Aspect 1)
A method of removing undesirable features 126 in an image comprising:
Forming a transformed image 140 by transforming the image from the spatial domain 136 to the frequency domain 138;
Forming a filtered image 154 in the frequency domain 138 by applying a filter 152 to the transformed image 140;
Returning the filtered image 154 from the frequency domain 138 to the spatial domain 136 to form a modified image 156 where the intensity of the unwanted feature 126 in the modified image 156 is undesirable in the original image. Increasing steps compared to the strength of the feature 126; and
Forming processed image 160 by removing unwanted features 126 from the image using modified image 156
Including methods.
(Aspect 2)
Forming the transformed image 140 by transforming the image from the spatial domain 136 to the frequency domain 138;
Computing a Fourier transform 142 of the image that transforms the image from the spatial domain 136 to the transformed image 140 in the frequency domain 138;
A method according to aspect 1, comprising:
(Aspect 3)
Forming a filtered image 154 in the frequency domain 138 by applying a filter 152 to the transformed image 140;
Forming a filtered image 154 by changing the amplitude component 144 of the image Fourier transform 142 without changing the phase component 146 of the image Fourier transform 142;
A method according to embodiment 2.
(Aspect 4)
Forming the filtered image 154 by changing the amplitude component 144 of the image Fourier transform 142 without changing the phase component 146 of the image Fourier transform 142;
In the step of forming a filtered image 154 by using the root filter 228 to scale the amplitude of the amplitude component 144 of the Fourier transform 142 of the image without affecting the phase of the phase component 146 of the Fourier transform 142. The route filter 228 includes scaling the amplitude nonlinearly;
A method according to aspect 3.
(Aspect 5)
Forming the modified image 156 by returning the filtered image 154 from the frequency domain 138 to the spatial domain 136;
Computing an inverse Fourier transform 142 of the filtered image 154 that transforms the filtered image 154 from the frequency domain 138 to the spatial domain 136, wherein the intensity of the undesired features 126 in the modified image 156 Increasing compared to the intensity of the undesired feature 126 of
A method according to aspect 1.
(Aspect 6)
Forming processed image 160 by removing undesirable features 126 from the image using modified image 156 includes the steps of:
Forming a threshold application image 242 by applying a threshold 240 to the modified image 156 and retaining pixels in the modified image 156 having a value above the threshold 240;
Convolution of the modified image 156 with the image point spread function 246 to form a convolved image, wherein the convolved image causes the second of the undesired features 126 in the original image to be The first position 158 of the undesired feature 126 in the convolved image corresponding to the position 158 is indicated, and
Subtracting the convolved image 248 from the image to form a processed image 160, wherein the image portion at the second position 158 of the undesired feature 126 in the image is removed from the image, Step of removing unwanted features 126
A method according to aspect 1, comprising:
(Aspect 7)
A vehicle inspection system,
A sensor system 104 configured to generate an image 106 of the vehicle;
Receiving an image from the sensor system 104; forming a transformed image 140 by transforming the image from the spatial domain 136 to the frequency domain 138; applying a filter 152 to the transformed image 140 to filter the image in the frequency domain 138 A modified image 156 is formed by returning the filtered image 154 from the frequency domain 138 back to the spatial domain 136, where the intensity of the undesired feature 126 in the modified image 156 is desirable in the original image. An image processor 120 configured to form a processed image 160 by removing the undesired features 126 from the image using the processed image 156; And
An image analyzer 122 configured to analyze the processed image 160 to determine if a mismatch exists in the vehicle.
Vehicle inspection system with
(Aspect 8)
In analyzing the processed image 160 to determine if a mismatch exists, the image analyzer 122 forms a comparison spectrum by comparing the processed image 160 with another image of the vehicle; 8. The vehicle inspection system of aspect 7, wherein the vehicle inspection system is used to determine whether a mismatch exists.

Claims (7)

A method for removing undesirable features ( 126 ) in an image comprising:
By converting the image into the frequency domain (138) from the spatial domain (136), forming a transformed image (140),
Forming a filtered image ( 154 ) in the frequency domain ( 138 ) by applying a filter ( 152 ) to the transformed image ( 140 ) ;
The filtered image (154) and then re-converted to the spatial domain (136) from the frequency domain (138), comprising the steps of forming a modified image (156), undesirable features of the modified image (156) in ( strength of 126), is increased compared with the strength of undesirable features in the image (126), the step, and
Step of removing the feature (126) undesired from the image,
Only including,
Removing unwanted features (126) comprises:
Applying a threshold (240) to the modified image (156) to form a threshold application image (242) by retaining pixels in the modified image (156) having a value above the threshold (240). ,
A step of forming a convolved image (248) by convolving the threshold application image (242) using a point spread function (246) of the image, wherein the convolved image (248) is the image Indicating a first position of the undesired feature (126) in the convolved image (248) corresponding to a second position of the undesired feature (126) in, and
Subtracting the convolved image (248) from the image to form a processed image (160), wherein the image portion at the second location of the undesired features (126) in the image is Removing undesired features (126) by removal from the image
Including a method. By converting the image into the frequency domain (138) from the spatial domain (136), the step of forming a transformed image (140),
Comprising the step of calculating a Fourier transform (142) of the image to convert the image from the spatial domain (136) to the transformed image (140) in the frequency domain (138), The method of claim 1. Forming a filtered image ( 154 ) in the frequency domain ( 138 ) by applying a filter ( 152 ) to the transformed image ( 140 ) ;
By changing the amplitude component (144) of the Fourier transform of the image without changing the phase component (146) (142) of the Fourier transform of the image (142), forming a filtered image (154) Including,
The method of claim 2. By changing the amplitude component (144) of the Fourier transform of the image without changing the phase component (146) (142) of the Fourier transform of the image (142), the step of forming the filtered image (154) ,
With root filter (228), to scale the amplitude of the amplitude component (144) of the Fourier transform of the image without affecting the phase of the phase component (146) (142) of the Fourier transform (142) To form a filtered image ( 154 ) , the root filter ( 228 ) non-linearly scaling the amplitude,
The method of claim 3. Forming the modified image ( 156 ) by re-transforming the filtered image ( 154 ) from the frequency domain ( 138 ) to the spatial domain ( 136 ) ,
Computing an inverse Fourier transform ( 142 ) of the filtered image ( 154 ) that transforms the filtered image ( 154 ) from the frequency domain ( 138 ) to the spatial domain ( 136 ) , in the modified image ( 156 ) intensity of undesirable features (126) comprises a step of increase compared to the intensity of the undesirable features (126) in said image,
The method of claim 1. A vehicle inspection system,
Configured sensor system to generate a vehicle image (106) (104),
Receiving an image from the sensor system ( 104 ) ; forming the transformed image ( 140 ) by transforming the image from the spatial domain ( 136 ) to the frequency domain ( 138 ); and applying a filter ( 152 ) to the transformed image ( 140 ) to form a filtered image (154) in the frequency domain (138) within by application to; corrected image by re-converting the filtered image (154) in the spatial domain (136) from the frequency domain (138) (156 preferably from and said image;) is formed, the strength is increased compared with the unwanted feature intensity in the image of the undesirable features in this case modified image (156) (126) (126) is by Uni configured to remove free features (126) image processing pro A Tsu service (120), corrected image (156) image processor (120 configured to form a removed to the processed image (160) features (126) undesired from the image by using the ),
A computer system (105) comprising:
Forming a threshold application image (242) by applying a threshold (240) to the modified image (156) and retaining pixels in the modified image (156) having a value above the threshold (240);
The threshold application image (242) is convolved with the image point spread function (246) to form a convolved image (248), wherein the convolved image (248) is preferably within the image. Indicating a first location of the undesired feature (126) in the convolved image (248) corresponding to a second location of the missing feature (126), and
Subtracting the convolved image (248) from the image forms a processed image (160), and image portions at the second location of undesirable features (126) in the image are removed from the image. The computer system (105) configured to remove undesirable features (126), as well as the processed image ( 160 ) to determine if a mismatch exists in the vehicle. Image analyzer configured as ( 122 )
Vehicle inspection system with By analyzing the processed image (160) in determining whether the mismatch is present, the image analyzer (122), a comparison by comparing the different images of the processed image (160) Vehicle formed; determining whether inconsistencies exist with comparison, vehicle inspection system according to claim 6.